[Docs] DeepSeek-V4 cookbook: drop --disable-flashinfer-autotune from GB300 Flash low-latency (#28590)

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
zijiexia
2026-06-18 00:05:21 -07:00
committed by GitHub
co-authored by Claude Opus 4.8
parent b7ae7149e8
commit 3f66873304
2 changed files with 7 additions and 1 deletions
@@ -140,6 +140,13 @@ export const benchmarks = [
// ==================================================================== // ====================================================================
{ {
match: { hw: "gb300", variant: "flash", quant: "fp4", strategy: "low-latency", nodes: "single" }, match: { hw: "gb300", variant: "flash", quant: "fp4", strategy: "low-latency", nodes: "single" },
sglang_version: "0.5.13.post1",
speed: [
{ workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 1 },
ttft_ms: 463, tpot_ms: 4.19, tokens_per_sec_per_gpu: 35 },
{ workload: { dataset: "random", isl: 8192, osl: 1024, max_concurrency: 16 },
ttft_ms: 436, tpot_ms: 8.93, tokens_per_sec_per_gpu: 336 },
],
}, },
{ {
match: { hw: "gb300", variant: "flash", quant: "fp4", strategy: "balanced", nodes: "single" }, match: { hw: "gb300", variant: "flash", quant: "fp4", strategy: "balanced", nodes: "single" },
@@ -743,7 +743,6 @@ sgl-eval run aime25 \\
"--speculative-eagle-topk 1", "--speculative-eagle-topk 1",
"--speculative-num-draft-tokens 4", "--speculative-num-draft-tokens 4",
"--chunked-prefill-size 4096", "--chunked-prefill-size 4096",
"--disable-flashinfer-autotune",
"--swa-full-tokens-ratio 0.1", "--swa-full-tokens-ratio 0.1",
"--host {{HOST_IP}}", "--host {{HOST_IP}}",
"--port {{PORT}}", "--port {{PORT}}",